As users have ever-increasing requirements on rates of data transmission, a development trend of base station miniaturization becomes increasingly obvious. Therefore, mobile communications networks are more dynamic, a quantity of network elements that need to be maintained by a carrier dramatically increases, and accordingly, maintenance expenditure that needs to be invested becomes increasingly high. In addition, high mobility of applications of a user leads to increasingly frequent changes in network services. A self-organizing network (“SON”) technology is proposed, attempting to implement automation as far as possible at planning, deployment, and operation and maintenance stages of a mobile communications network, so as adapt to network changes, and achieve an objective of reducing operation expenditure (“OPEX”).
For network performance optimization based on the SON technology, a model-based technical solution or a model-free technical solution may be used. The model-based technical solution is more suitable for actual application, but requires a model used for the network performance optimization to be very accurate, and to be capable of adapting to a service change of a mobile user.
Services of mobile users are increasingly diverse and dynamic. Generally, the services of the mobile users may be divided into a guaranteed bit rate (“GBR”) service and a non-GBR (“non-GBR”) service according to service requirements. A priority of the non-GBR service is lower than that of the GBR service, and a rate requirement of the non-GBR service is also uncertain. Therefore, the non-GBR service is more dynamic, and is difficult to predict.
Therefore, a key performance indicator (“KPI”) model that is usually used at present is based on load estimation, and the load estimation is based on a rate requirement of the GBR service, which causes the KPI model to be inaccurate. Therefore, a solution obtained during KPI optimization is not optimal, and performance of a communications network cannot be greatly improved.